Character recognition method, character recognition apparatus and recording medium on which a character recognition program is recorded

1. A character recognition method wherein a slant angle of a character is corrected to recognize the character, comprising the steps of: estimating a slant angle of a character; evaluating a validity of the slant angle estimated value, said evaluating step including calculating a validity evaluation value indicative of the validity of the slant angle estimated value; determining a slant correction amount from the slant angle estimated value and the validity evaluation value; performing slant correction of the character based on the slant correction amount, said performing step including correcting a slant of the character by a percentage of the slant angle estimated value, said percentage based on a magnitude of the validity evaluation value calculated in said evaluating step; and performing character recognition of the character.

2. A character recognition method as claimed in claim 1, wherein said evaluating step includes calculating the validity of the slant angle estimated value based on an information amount derived from evaluation values generated in said estimating step.

3. A character recognition method wherein a slant angle of a character is corrected to recognize the character, comprising the steps of: estimating a slant angle of a character; evaluating a validity of the slant angle estimated value; determining a slant correction amount from the slant angle estimated value and the validity evaluation value; performing slant correction of the character based on the slant correction amount and then performing character recognition of the character to extract character recognition candidates; and discriminating based on the validity evaluated value whether or not a stroke group in a character image has a fixed directionality to selectively adopt the character recognition candidates to reduce the number of candidates.

4. A character recognition method as claimed in claim 3, further comprising: determining a magnitude of the validity evaluated value, and wherein a degree of slant correction performed in said performing steps is dependent upon the magnitude of the validity evaluation value.

5. A character recognition method as claimed in claim 4, wherein said discriminating step includes: determining that said stroke group does not have said fixed directionality if the magnitude of the validity evaluated value lies within a predetermined range; and determining that said stroke group has said fixed directionality if the magnitude of the validity evaluated value does not lie within said predetermined range.

6. A character recognition method as claimed in claim 5, further comprising: removing candidates of kanji are removed from said character recognition candidates when the validity evaluated value lies within said predetermined range; and removing other than kanji candidates when the validity evaluated value does not lie within said predetermined range.

7. A character recognition apparatus wherein a slant angle of a character is corrected to recognize the character, comprising: character slant estimation means for calculating and outputting an estimated slant angle value of a character; estimated value evaluation means for calculating and outputting an evaluation value representative of a validity of the estimated slant angle value; and slant correction means for determining a character slant correction amount based on outputs of said character slant estimation means and said estimated value evaluation means to effect slant correction of the character, said slant correction means correcting a slant of the character by a percentage of the estimated slant angle value, said percentage based on a magnitude of the validity evaluation value calculated in said estimated value evaluation means.

8. A character recognition apparatus as claimed in claim 7, wherein said character slant estimation means sets a plurality of slant angle candidates, calculates evaluation values representative of degrees of correctness of the slant angle candidates, calculates a slant angle estimated value based on the evaluation values representative of the degrees of correctness, and outputs the slant angle estimated value together with the evaluation value representative of the degree of correctness.

9. A character recognition apparatus as claimed in claim 7, wherein, when the slant angle estimated value is to be calculated based on the evaluation values representative of the degrees of correctness, said character slant estimation means searches for the highest value of the evaluation values representative of the degrees of correctness and determines one of the slant angle candidates which corresponds to the highest value as the slant angle estimated value.

10. A character recognition apparatus as claimed in claim 7, wherein said estimated value evaluation means calculates and outputs an information amount that the entire evaluation values representative of the degrees of correctness outputted from said character slant estimation means has as an evaluation value representative of a validity of the slant estimated value outputted from said character slant estimation means.

11. A character recognition apparatus as claimed in claim 7, wherein said estimated value evaluation means calculates and outputs a square sum, a square integral or a variance of the evaluated values indicative of the degrees of correctness outputted from said character slant estimation means as the evaluation value representative of the validity of the slant estimated value outputted from said character slant estimation means.

12. A character recognition apparatus as claimed in claim 7, wherein said slant correction means normalizes the evaluation value outputted from said estimated value evaluation means and indicative of the validity to a value from 0 to 1 using a monotone non-decreasing function and utilizes the normalized value as an indication of said percentage for adjustment of the magnitude of the slant correction amount.

13. A character recognition apparatus as claimed in claim 12, wherein a Sigmoid function is used as the monotone non-decreasing function.

14. A character recognition apparatus as claimed in claim 12, wherein a linear function or a sectional linear function is used as the monotone non-decreasing function.

15. A character recognition apparatus as claimed in claim 7, further comprising recognition candidate reduction means for discriminating based on the output of said estimated value evaluation means whether or not a stroke group in a character image has a fixed directionality to selectively adopt the character recognition candidates to reduce the number of candidates.

16. A character recognition apparatus as claimed in claim 15, wherein said recognition candidate reduction means performs discrimination between kanji and non-kanji based on the output of said estimated value evaluation means.

17. A recording medium on which a recognition program which operates on a computer is recorded, the character recognition program causing the computer to execute the steps of: estimating a slant angle of a character; evaluating a validity of the slant angle estimated value, said evaluating step including calculating a validity evaluation value indicative of the validity of the slant angle estimated value; determining a slant correction amount from the slant angle estimated value and the validity evaluation value; performing slant correction of the character based on the slant correction amount, said performing step including correcting a slant of the character by a percentage of the slant angle estimated value, said percentage based on a magnitude of the validity evaluation value calculated in said evaluating step; and performing character recognition of the character.